A near-surface groundwater prospectivity model for the Main Karoo Basin of South Africa derived from multivariate machine learning

Climate change affecting arid and semi-arid regions increases the periodicity and intensity of droughts resulting in a need for the development of effective groundwater exploration techniques. Here, the availability of near-surface groundwater in the Main Karoo Basin (MKB) is evaluated using multivariate machine learning models. These models integrate 21 conditioning factors ranging from spectral indices, topographical features, geological formations, and hydrological parameters. Among the five machine learning (ML) models tested, the Fast Tree Decision Learning models achieved the highest classification accuracy (81.4%) and a robust Receiver Operating Characteristic (ROC) area curve of 0.87. The resultant near-surface groundwater prospectivity model showed a statistically significant (p < 0.00001) alignment with the spatial locations of high-yielding boreholes, springs, and groundwater-dependent vegetation. Areas with a high potential for near-surface groundwater were identified along the Drakensberg Escarpment, the Cape Fold Belt, and along the eastern MKB adjacent to the Indian Ocean. In the arid western MKB, localized zones identified to be highly prospective for near-surface groundwater coincide with the intersections of drainage networks and major geological structures. Geo-hydrologically, these areas are characterized by borehole yields exceeding 9 L/s. This study illustrates the effectiveness the ML models that harness regional datasets in characterizing prospective areas for near-surface groundwater in data-scarce, arid environments.